xref: /linux/net/ipv6/ipv6_sockglue.c (revision 08ec212c0f92cbf30e3ecc7349f18151714041d6)
1 /*
2  *	IPv6 BSD socket options interface
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Pedro Roque		<roque@di.fc.ul.pt>
7  *
8  *	Based on linux/net/ipv4/ip_sockglue.c
9  *
10  *	This program is free software; you can redistribute it and/or
11  *      modify it under the terms of the GNU General Public License
12  *      as published by the Free Software Foundation; either version
13  *      2 of the License, or (at your option) any later version.
14  *
15  *	FIXME: Make the setsockopt code POSIX compliant: That is
16  *
17  *	o	Truncate getsockopt returns
18  *	o	Return an optlen of the truncated length if need be
19  *
20  *	Changes:
21  *	David L Stevens <dlstevens@us.ibm.com>:
22  *		- added multicast source filtering API for MLDv2
23  */
24 
25 #include <linux/module.h>
26 #include <linux/capability.h>
27 #include <linux/errno.h>
28 #include <linux/types.h>
29 #include <linux/socket.h>
30 #include <linux/sockios.h>
31 #include <linux/net.h>
32 #include <linux/in6.h>
33 #include <linux/mroute6.h>
34 #include <linux/netdevice.h>
35 #include <linux/if_arp.h>
36 #include <linux/init.h>
37 #include <linux/sysctl.h>
38 #include <linux/netfilter.h>
39 #include <linux/slab.h>
40 
41 #include <net/sock.h>
42 #include <net/snmp.h>
43 #include <net/ipv6.h>
44 #include <net/ndisc.h>
45 #include <net/protocol.h>
46 #include <net/transp_v6.h>
47 #include <net/ip6_route.h>
48 #include <net/addrconf.h>
49 #include <net/inet_common.h>
50 #include <net/tcp.h>
51 #include <net/udp.h>
52 #include <net/udplite.h>
53 #include <net/xfrm.h>
54 #include <net/compat.h>
55 
56 #include <asm/uaccess.h>
57 
58 struct ip6_ra_chain *ip6_ra_chain;
59 DEFINE_RWLOCK(ip6_ra_lock);
60 
61 int ip6_ra_control(struct sock *sk, int sel)
62 {
63 	struct ip6_ra_chain *ra, *new_ra, **rap;
64 
65 	/* RA packet may be delivered ONLY to IPPROTO_RAW socket */
66 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num != IPPROTO_RAW)
67 		return -ENOPROTOOPT;
68 
69 	new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
70 
71 	write_lock_bh(&ip6_ra_lock);
72 	for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
73 		if (ra->sk == sk) {
74 			if (sel>=0) {
75 				write_unlock_bh(&ip6_ra_lock);
76 				kfree(new_ra);
77 				return -EADDRINUSE;
78 			}
79 
80 			*rap = ra->next;
81 			write_unlock_bh(&ip6_ra_lock);
82 
83 			sock_put(sk);
84 			kfree(ra);
85 			return 0;
86 		}
87 	}
88 	if (new_ra == NULL) {
89 		write_unlock_bh(&ip6_ra_lock);
90 		return -ENOBUFS;
91 	}
92 	new_ra->sk = sk;
93 	new_ra->sel = sel;
94 	new_ra->next = ra;
95 	*rap = new_ra;
96 	sock_hold(sk);
97 	write_unlock_bh(&ip6_ra_lock);
98 	return 0;
99 }
100 
101 static
102 struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
103 					   struct ipv6_txoptions *opt)
104 {
105 	if (inet_sk(sk)->is_icsk) {
106 		if (opt &&
107 		    !((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
108 		    inet_sk(sk)->inet_daddr != LOOPBACK4_IPV6) {
109 			struct inet_connection_sock *icsk = inet_csk(sk);
110 			icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
111 			icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
112 		}
113 		opt = xchg(&inet6_sk(sk)->opt, opt);
114 	} else {
115 		spin_lock(&sk->sk_dst_lock);
116 		opt = xchg(&inet6_sk(sk)->opt, opt);
117 		spin_unlock(&sk->sk_dst_lock);
118 	}
119 	sk_dst_reset(sk);
120 
121 	return opt;
122 }
123 
124 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
125 		    char __user *optval, unsigned int optlen)
126 {
127 	struct ipv6_pinfo *np = inet6_sk(sk);
128 	struct net *net = sock_net(sk);
129 	int val, valbool;
130 	int retv = -ENOPROTOOPT;
131 
132 	if (optval == NULL)
133 		val=0;
134 	else {
135 		if (optlen >= sizeof(int)) {
136 			if (get_user(val, (int __user *) optval))
137 				return -EFAULT;
138 		} else
139 			val = 0;
140 	}
141 
142 	valbool = (val!=0);
143 
144 	if (ip6_mroute_opt(optname))
145 		return ip6_mroute_setsockopt(sk, optname, optval, optlen);
146 
147 	lock_sock(sk);
148 
149 	switch (optname) {
150 
151 	case IPV6_ADDRFORM:
152 		if (optlen < sizeof(int))
153 			goto e_inval;
154 		if (val == PF_INET) {
155 			struct ipv6_txoptions *opt;
156 			struct sk_buff *pktopt;
157 
158 			if (sk->sk_type == SOCK_RAW)
159 				break;
160 
161 			if (sk->sk_protocol == IPPROTO_UDP ||
162 			    sk->sk_protocol == IPPROTO_UDPLITE) {
163 				struct udp_sock *up = udp_sk(sk);
164 				if (up->pending == AF_INET6) {
165 					retv = -EBUSY;
166 					break;
167 				}
168 			} else if (sk->sk_protocol != IPPROTO_TCP)
169 				break;
170 
171 			if (sk->sk_state != TCP_ESTABLISHED) {
172 				retv = -ENOTCONN;
173 				break;
174 			}
175 
176 			if (ipv6_only_sock(sk) ||
177 			    !ipv6_addr_v4mapped(&np->daddr)) {
178 				retv = -EADDRNOTAVAIL;
179 				break;
180 			}
181 
182 			fl6_free_socklist(sk);
183 			ipv6_sock_mc_close(sk);
184 
185 			/*
186 			 * Sock is moving from IPv6 to IPv4 (sk_prot), so
187 			 * remove it from the refcnt debug socks count in the
188 			 * original family...
189 			 */
190 			sk_refcnt_debug_dec(sk);
191 
192 			if (sk->sk_protocol == IPPROTO_TCP) {
193 				struct inet_connection_sock *icsk = inet_csk(sk);
194 				local_bh_disable();
195 				sock_prot_inuse_add(net, sk->sk_prot, -1);
196 				sock_prot_inuse_add(net, &tcp_prot, 1);
197 				local_bh_enable();
198 				sk->sk_prot = &tcp_prot;
199 				icsk->icsk_af_ops = &ipv4_specific;
200 				sk->sk_socket->ops = &inet_stream_ops;
201 				sk->sk_family = PF_INET;
202 				tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
203 			} else {
204 				struct proto *prot = &udp_prot;
205 
206 				if (sk->sk_protocol == IPPROTO_UDPLITE)
207 					prot = &udplite_prot;
208 				local_bh_disable();
209 				sock_prot_inuse_add(net, sk->sk_prot, -1);
210 				sock_prot_inuse_add(net, prot, 1);
211 				local_bh_enable();
212 				sk->sk_prot = prot;
213 				sk->sk_socket->ops = &inet_dgram_ops;
214 				sk->sk_family = PF_INET;
215 			}
216 			opt = xchg(&np->opt, NULL);
217 			if (opt)
218 				sock_kfree_s(sk, opt, opt->tot_len);
219 			pktopt = xchg(&np->pktoptions, NULL);
220 			kfree_skb(pktopt);
221 
222 			sk->sk_destruct = inet_sock_destruct;
223 			/*
224 			 * ... and add it to the refcnt debug socks count
225 			 * in the new family. -acme
226 			 */
227 			sk_refcnt_debug_inc(sk);
228 			module_put(THIS_MODULE);
229 			retv = 0;
230 			break;
231 		}
232 		goto e_inval;
233 
234 	case IPV6_V6ONLY:
235 		if (optlen < sizeof(int) ||
236 		    inet_sk(sk)->inet_num)
237 			goto e_inval;
238 		np->ipv6only = valbool;
239 		retv = 0;
240 		break;
241 
242 	case IPV6_RECVPKTINFO:
243 		if (optlen < sizeof(int))
244 			goto e_inval;
245 		np->rxopt.bits.rxinfo = valbool;
246 		retv = 0;
247 		break;
248 
249 	case IPV6_2292PKTINFO:
250 		if (optlen < sizeof(int))
251 			goto e_inval;
252 		np->rxopt.bits.rxoinfo = valbool;
253 		retv = 0;
254 		break;
255 
256 	case IPV6_RECVHOPLIMIT:
257 		if (optlen < sizeof(int))
258 			goto e_inval;
259 		np->rxopt.bits.rxhlim = valbool;
260 		retv = 0;
261 		break;
262 
263 	case IPV6_2292HOPLIMIT:
264 		if (optlen < sizeof(int))
265 			goto e_inval;
266 		np->rxopt.bits.rxohlim = valbool;
267 		retv = 0;
268 		break;
269 
270 	case IPV6_RECVRTHDR:
271 		if (optlen < sizeof(int))
272 			goto e_inval;
273 		np->rxopt.bits.srcrt = valbool;
274 		retv = 0;
275 		break;
276 
277 	case IPV6_2292RTHDR:
278 		if (optlen < sizeof(int))
279 			goto e_inval;
280 		np->rxopt.bits.osrcrt = valbool;
281 		retv = 0;
282 		break;
283 
284 	case IPV6_RECVHOPOPTS:
285 		if (optlen < sizeof(int))
286 			goto e_inval;
287 		np->rxopt.bits.hopopts = valbool;
288 		retv = 0;
289 		break;
290 
291 	case IPV6_2292HOPOPTS:
292 		if (optlen < sizeof(int))
293 			goto e_inval;
294 		np->rxopt.bits.ohopopts = valbool;
295 		retv = 0;
296 		break;
297 
298 	case IPV6_RECVDSTOPTS:
299 		if (optlen < sizeof(int))
300 			goto e_inval;
301 		np->rxopt.bits.dstopts = valbool;
302 		retv = 0;
303 		break;
304 
305 	case IPV6_2292DSTOPTS:
306 		if (optlen < sizeof(int))
307 			goto e_inval;
308 		np->rxopt.bits.odstopts = valbool;
309 		retv = 0;
310 		break;
311 
312 	case IPV6_TCLASS:
313 		if (optlen < sizeof(int))
314 			goto e_inval;
315 		if (val < -1 || val > 0xff)
316 			goto e_inval;
317 		/* RFC 3542, 6.5: default traffic class of 0x0 */
318 		if (val == -1)
319 			val = 0;
320 		np->tclass = val;
321 		retv = 0;
322 		break;
323 
324 	case IPV6_RECVTCLASS:
325 		if (optlen < sizeof(int))
326 			goto e_inval;
327 		np->rxopt.bits.rxtclass = valbool;
328 		retv = 0;
329 		break;
330 
331 	case IPV6_FLOWINFO:
332 		if (optlen < sizeof(int))
333 			goto e_inval;
334 		np->rxopt.bits.rxflow = valbool;
335 		retv = 0;
336 		break;
337 
338 	case IPV6_RECVPATHMTU:
339 		if (optlen < sizeof(int))
340 			goto e_inval;
341 		np->rxopt.bits.rxpmtu = valbool;
342 		retv = 0;
343 		break;
344 
345 	case IPV6_TRANSPARENT:
346 		if (valbool && !capable(CAP_NET_ADMIN) && !capable(CAP_NET_RAW)) {
347 			retv = -EPERM;
348 			break;
349 		}
350 		if (optlen < sizeof(int))
351 			goto e_inval;
352 		/* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
353 		inet_sk(sk)->transparent = valbool;
354 		retv = 0;
355 		break;
356 
357 	case IPV6_RECVORIGDSTADDR:
358 		if (optlen < sizeof(int))
359 			goto e_inval;
360 		np->rxopt.bits.rxorigdstaddr = valbool;
361 		retv = 0;
362 		break;
363 
364 	case IPV6_HOPOPTS:
365 	case IPV6_RTHDRDSTOPTS:
366 	case IPV6_RTHDR:
367 	case IPV6_DSTOPTS:
368 	{
369 		struct ipv6_txoptions *opt;
370 
371 		/* remove any sticky options header with a zero option
372 		 * length, per RFC3542.
373 		 */
374 		if (optlen == 0)
375 			optval = NULL;
376 		else if (optval == NULL)
377 			goto e_inval;
378 		else if (optlen < sizeof(struct ipv6_opt_hdr) ||
379 			 optlen & 0x7 || optlen > 8 * 255)
380 			goto e_inval;
381 
382 		/* hop-by-hop / destination options are privileged option */
383 		retv = -EPERM;
384 		if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
385 			break;
386 
387 		opt = ipv6_renew_options(sk, np->opt, optname,
388 					 (struct ipv6_opt_hdr __user *)optval,
389 					 optlen);
390 		if (IS_ERR(opt)) {
391 			retv = PTR_ERR(opt);
392 			break;
393 		}
394 
395 		/* routing header option needs extra check */
396 		retv = -EINVAL;
397 		if (optname == IPV6_RTHDR && opt && opt->srcrt) {
398 			struct ipv6_rt_hdr *rthdr = opt->srcrt;
399 			switch (rthdr->type) {
400 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
401 			case IPV6_SRCRT_TYPE_2:
402 				if (rthdr->hdrlen != 2 ||
403 				    rthdr->segments_left != 1)
404 					goto sticky_done;
405 
406 				break;
407 #endif
408 			default:
409 				goto sticky_done;
410 			}
411 		}
412 
413 		retv = 0;
414 		opt = ipv6_update_options(sk, opt);
415 sticky_done:
416 		if (opt)
417 			sock_kfree_s(sk, opt, opt->tot_len);
418 		break;
419 	}
420 
421 	case IPV6_PKTINFO:
422 	{
423 		struct in6_pktinfo pkt;
424 
425 		if (optlen == 0)
426 			goto e_inval;
427 		else if (optlen < sizeof(struct in6_pktinfo) || optval == NULL)
428 			goto e_inval;
429 
430 		if (copy_from_user(&pkt, optval, sizeof(struct in6_pktinfo))) {
431 				retv = -EFAULT;
432 				break;
433 		}
434 		if (sk->sk_bound_dev_if && pkt.ipi6_ifindex != sk->sk_bound_dev_if)
435 			goto e_inval;
436 
437 		np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
438 		np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr;
439 		retv = 0;
440 		break;
441 	}
442 
443 	case IPV6_2292PKTOPTIONS:
444 	{
445 		struct ipv6_txoptions *opt = NULL;
446 		struct msghdr msg;
447 		struct flowi6 fl6;
448 		int junk;
449 
450 		memset(&fl6, 0, sizeof(fl6));
451 		fl6.flowi6_oif = sk->sk_bound_dev_if;
452 		fl6.flowi6_mark = sk->sk_mark;
453 
454 		if (optlen == 0)
455 			goto update;
456 
457 		/* 1K is probably excessive
458 		 * 1K is surely not enough, 2K per standard header is 16K.
459 		 */
460 		retv = -EINVAL;
461 		if (optlen > 64*1024)
462 			break;
463 
464 		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
465 		retv = -ENOBUFS;
466 		if (opt == NULL)
467 			break;
468 
469 		memset(opt, 0, sizeof(*opt));
470 		opt->tot_len = sizeof(*opt) + optlen;
471 		retv = -EFAULT;
472 		if (copy_from_user(opt+1, optval, optlen))
473 			goto done;
474 
475 		msg.msg_controllen = optlen;
476 		msg.msg_control = (void*)(opt+1);
477 
478 		retv = datagram_send_ctl(net, sk, &msg, &fl6, opt, &junk, &junk,
479 					 &junk);
480 		if (retv)
481 			goto done;
482 update:
483 		retv = 0;
484 		opt = ipv6_update_options(sk, opt);
485 done:
486 		if (opt)
487 			sock_kfree_s(sk, opt, opt->tot_len);
488 		break;
489 	}
490 	case IPV6_UNICAST_HOPS:
491 		if (optlen < sizeof(int))
492 			goto e_inval;
493 		if (val > 255 || val < -1)
494 			goto e_inval;
495 		np->hop_limit = val;
496 		retv = 0;
497 		break;
498 
499 	case IPV6_MULTICAST_HOPS:
500 		if (sk->sk_type == SOCK_STREAM)
501 			break;
502 		if (optlen < sizeof(int))
503 			goto e_inval;
504 		if (val > 255 || val < -1)
505 			goto e_inval;
506 		np->mcast_hops = (val == -1 ? IPV6_DEFAULT_MCASTHOPS : val);
507 		retv = 0;
508 		break;
509 
510 	case IPV6_MULTICAST_LOOP:
511 		if (optlen < sizeof(int))
512 			goto e_inval;
513 		if (val != valbool)
514 			goto e_inval;
515 		np->mc_loop = valbool;
516 		retv = 0;
517 		break;
518 
519 	case IPV6_UNICAST_IF:
520 	{
521 		struct net_device *dev = NULL;
522 		int ifindex;
523 
524 		if (optlen != sizeof(int))
525 			goto e_inval;
526 
527 		ifindex = (__force int)ntohl((__force __be32)val);
528 		if (ifindex == 0) {
529 			np->ucast_oif = 0;
530 			retv = 0;
531 			break;
532 		}
533 
534 		dev = dev_get_by_index(net, ifindex);
535 		retv = -EADDRNOTAVAIL;
536 		if (!dev)
537 			break;
538 		dev_put(dev);
539 
540 		retv = -EINVAL;
541 		if (sk->sk_bound_dev_if)
542 			break;
543 
544 		np->ucast_oif = ifindex;
545 		retv = 0;
546 		break;
547 	}
548 
549 	case IPV6_MULTICAST_IF:
550 		if (sk->sk_type == SOCK_STREAM)
551 			break;
552 		if (optlen < sizeof(int))
553 			goto e_inval;
554 
555 		if (val) {
556 			struct net_device *dev;
557 
558 			if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
559 				goto e_inval;
560 
561 			dev = dev_get_by_index(net, val);
562 			if (!dev) {
563 				retv = -ENODEV;
564 				break;
565 			}
566 			dev_put(dev);
567 		}
568 		np->mcast_oif = val;
569 		retv = 0;
570 		break;
571 	case IPV6_ADD_MEMBERSHIP:
572 	case IPV6_DROP_MEMBERSHIP:
573 	{
574 		struct ipv6_mreq mreq;
575 
576 		if (optlen < sizeof(struct ipv6_mreq))
577 			goto e_inval;
578 
579 		retv = -EPROTO;
580 		if (inet_sk(sk)->is_icsk)
581 			break;
582 
583 		retv = -EFAULT;
584 		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
585 			break;
586 
587 		if (optname == IPV6_ADD_MEMBERSHIP)
588 			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
589 		else
590 			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
591 		break;
592 	}
593 	case IPV6_JOIN_ANYCAST:
594 	case IPV6_LEAVE_ANYCAST:
595 	{
596 		struct ipv6_mreq mreq;
597 
598 		if (optlen < sizeof(struct ipv6_mreq))
599 			goto e_inval;
600 
601 		retv = -EFAULT;
602 		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
603 			break;
604 
605 		if (optname == IPV6_JOIN_ANYCAST)
606 			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
607 		else
608 			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
609 		break;
610 	}
611 	case MCAST_JOIN_GROUP:
612 	case MCAST_LEAVE_GROUP:
613 	{
614 		struct group_req greq;
615 		struct sockaddr_in6 *psin6;
616 
617 		if (optlen < sizeof(struct group_req))
618 			goto e_inval;
619 
620 		retv = -EFAULT;
621 		if (copy_from_user(&greq, optval, sizeof(struct group_req)))
622 			break;
623 		if (greq.gr_group.ss_family != AF_INET6) {
624 			retv = -EADDRNOTAVAIL;
625 			break;
626 		}
627 		psin6 = (struct sockaddr_in6 *)&greq.gr_group;
628 		if (optname == MCAST_JOIN_GROUP)
629 			retv = ipv6_sock_mc_join(sk, greq.gr_interface,
630 				&psin6->sin6_addr);
631 		else
632 			retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
633 				&psin6->sin6_addr);
634 		break;
635 	}
636 	case MCAST_JOIN_SOURCE_GROUP:
637 	case MCAST_LEAVE_SOURCE_GROUP:
638 	case MCAST_BLOCK_SOURCE:
639 	case MCAST_UNBLOCK_SOURCE:
640 	{
641 		struct group_source_req greqs;
642 		int omode, add;
643 
644 		if (optlen < sizeof(struct group_source_req))
645 			goto e_inval;
646 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
647 			retv = -EFAULT;
648 			break;
649 		}
650 		if (greqs.gsr_group.ss_family != AF_INET6 ||
651 		    greqs.gsr_source.ss_family != AF_INET6) {
652 			retv = -EADDRNOTAVAIL;
653 			break;
654 		}
655 		if (optname == MCAST_BLOCK_SOURCE) {
656 			omode = MCAST_EXCLUDE;
657 			add = 1;
658 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
659 			omode = MCAST_EXCLUDE;
660 			add = 0;
661 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
662 			struct sockaddr_in6 *psin6;
663 
664 			psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
665 			retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
666 				&psin6->sin6_addr);
667 			/* prior join w/ different source is ok */
668 			if (retv && retv != -EADDRINUSE)
669 				break;
670 			omode = MCAST_INCLUDE;
671 			add = 1;
672 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
673 			omode = MCAST_INCLUDE;
674 			add = 0;
675 		}
676 		retv = ip6_mc_source(add, omode, sk, &greqs);
677 		break;
678 	}
679 	case MCAST_MSFILTER:
680 	{
681 		struct group_filter *gsf;
682 
683 		if (optlen < GROUP_FILTER_SIZE(0))
684 			goto e_inval;
685 		if (optlen > sysctl_optmem_max) {
686 			retv = -ENOBUFS;
687 			break;
688 		}
689 		gsf = kmalloc(optlen,GFP_KERNEL);
690 		if (!gsf) {
691 			retv = -ENOBUFS;
692 			break;
693 		}
694 		retv = -EFAULT;
695 		if (copy_from_user(gsf, optval, optlen)) {
696 			kfree(gsf);
697 			break;
698 		}
699 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
700 		if (gsf->gf_numsrc >= 0x1ffffffU ||
701 		    gsf->gf_numsrc > sysctl_mld_max_msf) {
702 			kfree(gsf);
703 			retv = -ENOBUFS;
704 			break;
705 		}
706 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
707 			kfree(gsf);
708 			retv = -EINVAL;
709 			break;
710 		}
711 		retv = ip6_mc_msfilter(sk, gsf);
712 		kfree(gsf);
713 
714 		break;
715 	}
716 	case IPV6_ROUTER_ALERT:
717 		if (optlen < sizeof(int))
718 			goto e_inval;
719 		retv = ip6_ra_control(sk, val);
720 		break;
721 	case IPV6_MTU_DISCOVER:
722 		if (optlen < sizeof(int))
723 			goto e_inval;
724 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE)
725 			goto e_inval;
726 		np->pmtudisc = val;
727 		retv = 0;
728 		break;
729 	case IPV6_MTU:
730 		if (optlen < sizeof(int))
731 			goto e_inval;
732 		if (val && val < IPV6_MIN_MTU)
733 			goto e_inval;
734 		np->frag_size = val;
735 		retv = 0;
736 		break;
737 	case IPV6_RECVERR:
738 		if (optlen < sizeof(int))
739 			goto e_inval;
740 		np->recverr = valbool;
741 		if (!val)
742 			skb_queue_purge(&sk->sk_error_queue);
743 		retv = 0;
744 		break;
745 	case IPV6_FLOWINFO_SEND:
746 		if (optlen < sizeof(int))
747 			goto e_inval;
748 		np->sndflow = valbool;
749 		retv = 0;
750 		break;
751 	case IPV6_FLOWLABEL_MGR:
752 		retv = ipv6_flowlabel_opt(sk, optval, optlen);
753 		break;
754 	case IPV6_IPSEC_POLICY:
755 	case IPV6_XFRM_POLICY:
756 		retv = -EPERM;
757 		if (!capable(CAP_NET_ADMIN))
758 			break;
759 		retv = xfrm_user_policy(sk, optname, optval, optlen);
760 		break;
761 
762 	case IPV6_ADDR_PREFERENCES:
763 	    {
764 		unsigned int pref = 0;
765 		unsigned int prefmask = ~0;
766 
767 		if (optlen < sizeof(int))
768 			goto e_inval;
769 
770 		retv = -EINVAL;
771 
772 		/* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
773 		switch (val & (IPV6_PREFER_SRC_PUBLIC|
774 			       IPV6_PREFER_SRC_TMP|
775 			       IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
776 		case IPV6_PREFER_SRC_PUBLIC:
777 			pref |= IPV6_PREFER_SRC_PUBLIC;
778 			break;
779 		case IPV6_PREFER_SRC_TMP:
780 			pref |= IPV6_PREFER_SRC_TMP;
781 			break;
782 		case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
783 			break;
784 		case 0:
785 			goto pref_skip_pubtmp;
786 		default:
787 			goto e_inval;
788 		}
789 
790 		prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
791 			      IPV6_PREFER_SRC_TMP);
792 pref_skip_pubtmp:
793 
794 		/* check HOME/COA conflicts */
795 		switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
796 		case IPV6_PREFER_SRC_HOME:
797 			break;
798 		case IPV6_PREFER_SRC_COA:
799 			pref |= IPV6_PREFER_SRC_COA;
800 		case 0:
801 			goto pref_skip_coa;
802 		default:
803 			goto e_inval;
804 		}
805 
806 		prefmask &= ~IPV6_PREFER_SRC_COA;
807 pref_skip_coa:
808 
809 		/* check CGA/NONCGA conflicts */
810 		switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
811 		case IPV6_PREFER_SRC_CGA:
812 		case IPV6_PREFER_SRC_NONCGA:
813 		case 0:
814 			break;
815 		default:
816 			goto e_inval;
817 		}
818 
819 		np->srcprefs = (np->srcprefs & prefmask) | pref;
820 		retv = 0;
821 
822 		break;
823 	    }
824 	case IPV6_MINHOPCOUNT:
825 		if (optlen < sizeof(int))
826 			goto e_inval;
827 		if (val < 0 || val > 255)
828 			goto e_inval;
829 		np->min_hopcount = val;
830 		break;
831 	case IPV6_DONTFRAG:
832 		np->dontfrag = valbool;
833 		retv = 0;
834 		break;
835 	}
836 
837 	release_sock(sk);
838 
839 	return retv;
840 
841 e_inval:
842 	release_sock(sk);
843 	return -EINVAL;
844 }
845 
846 int ipv6_setsockopt(struct sock *sk, int level, int optname,
847 		    char __user *optval, unsigned int optlen)
848 {
849 	int err;
850 
851 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
852 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
853 
854 	if (level != SOL_IPV6)
855 		return -ENOPROTOOPT;
856 
857 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
858 #ifdef CONFIG_NETFILTER
859 	/* we need to exclude all possible ENOPROTOOPTs except default case */
860 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
861 			optname != IPV6_XFRM_POLICY) {
862 		lock_sock(sk);
863 		err = nf_setsockopt(sk, PF_INET6, optname, optval,
864 				optlen);
865 		release_sock(sk);
866 	}
867 #endif
868 	return err;
869 }
870 
871 EXPORT_SYMBOL(ipv6_setsockopt);
872 
873 #ifdef CONFIG_COMPAT
874 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
875 			   char __user *optval, unsigned int optlen)
876 {
877 	int err;
878 
879 	if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
880 		if (udp_prot.compat_setsockopt != NULL)
881 			return udp_prot.compat_setsockopt(sk, level, optname,
882 							  optval, optlen);
883 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
884 	}
885 
886 	if (level != SOL_IPV6)
887 		return -ENOPROTOOPT;
888 
889 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
890 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
891 			ipv6_setsockopt);
892 
893 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
894 #ifdef CONFIG_NETFILTER
895 	/* we need to exclude all possible ENOPROTOOPTs except default case */
896 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
897 	    optname != IPV6_XFRM_POLICY) {
898 		lock_sock(sk);
899 		err = compat_nf_setsockopt(sk, PF_INET6, optname,
900 					   optval, optlen);
901 		release_sock(sk);
902 	}
903 #endif
904 	return err;
905 }
906 
907 EXPORT_SYMBOL(compat_ipv6_setsockopt);
908 #endif
909 
910 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
911 				  int optname, char __user *optval, int len)
912 {
913 	struct ipv6_opt_hdr *hdr;
914 
915 	if (!opt)
916 		return 0;
917 
918 	switch(optname) {
919 	case IPV6_HOPOPTS:
920 		hdr = opt->hopopt;
921 		break;
922 	case IPV6_RTHDRDSTOPTS:
923 		hdr = opt->dst0opt;
924 		break;
925 	case IPV6_RTHDR:
926 		hdr = (struct ipv6_opt_hdr *)opt->srcrt;
927 		break;
928 	case IPV6_DSTOPTS:
929 		hdr = opt->dst1opt;
930 		break;
931 	default:
932 		return -EINVAL;	/* should not happen */
933 	}
934 
935 	if (!hdr)
936 		return 0;
937 
938 	len = min_t(unsigned int, len, ipv6_optlen(hdr));
939 	if (copy_to_user(optval, hdr, len))
940 		return -EFAULT;
941 	return len;
942 }
943 
944 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
945 		    char __user *optval, int __user *optlen, unsigned int flags)
946 {
947 	struct ipv6_pinfo *np = inet6_sk(sk);
948 	int len;
949 	int val;
950 
951 	if (ip6_mroute_opt(optname))
952 		return ip6_mroute_getsockopt(sk, optname, optval, optlen);
953 
954 	if (get_user(len, optlen))
955 		return -EFAULT;
956 	switch (optname) {
957 	case IPV6_ADDRFORM:
958 		if (sk->sk_protocol != IPPROTO_UDP &&
959 		    sk->sk_protocol != IPPROTO_UDPLITE &&
960 		    sk->sk_protocol != IPPROTO_TCP)
961 			return -ENOPROTOOPT;
962 		if (sk->sk_state != TCP_ESTABLISHED)
963 			return -ENOTCONN;
964 		val = sk->sk_family;
965 		break;
966 	case MCAST_MSFILTER:
967 	{
968 		struct group_filter gsf;
969 		int err;
970 
971 		if (len < GROUP_FILTER_SIZE(0))
972 			return -EINVAL;
973 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
974 			return -EFAULT;
975 		if (gsf.gf_group.ss_family != AF_INET6)
976 			return -EADDRNOTAVAIL;
977 		lock_sock(sk);
978 		err = ip6_mc_msfget(sk, &gsf,
979 			(struct group_filter __user *)optval, optlen);
980 		release_sock(sk);
981 		return err;
982 	}
983 
984 	case IPV6_2292PKTOPTIONS:
985 	{
986 		struct msghdr msg;
987 		struct sk_buff *skb;
988 
989 		if (sk->sk_type != SOCK_STREAM)
990 			return -ENOPROTOOPT;
991 
992 		msg.msg_control = optval;
993 		msg.msg_controllen = len;
994 		msg.msg_flags = flags;
995 
996 		lock_sock(sk);
997 		skb = np->pktoptions;
998 		if (skb)
999 			atomic_inc(&skb->users);
1000 		release_sock(sk);
1001 
1002 		if (skb) {
1003 			int err = datagram_recv_ctl(sk, &msg, skb);
1004 			kfree_skb(skb);
1005 			if (err)
1006 				return err;
1007 		} else {
1008 			if (np->rxopt.bits.rxinfo) {
1009 				struct in6_pktinfo src_info;
1010 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1011 					np->sticky_pktinfo.ipi6_ifindex;
1012 				src_info.ipi6_addr = np->mcast_oif ? np->daddr : np->sticky_pktinfo.ipi6_addr;
1013 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
1014 			}
1015 			if (np->rxopt.bits.rxhlim) {
1016 				int hlim = np->mcast_hops;
1017 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
1018 			}
1019 			if (np->rxopt.bits.rxtclass) {
1020 				int tclass = np->rcv_tclass;
1021 				put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
1022 			}
1023 			if (np->rxopt.bits.rxoinfo) {
1024 				struct in6_pktinfo src_info;
1025 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1026 					np->sticky_pktinfo.ipi6_ifindex;
1027 				src_info.ipi6_addr = np->mcast_oif ? np->daddr : np->sticky_pktinfo.ipi6_addr;
1028 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1029 			}
1030 			if (np->rxopt.bits.rxohlim) {
1031 				int hlim = np->mcast_hops;
1032 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1033 			}
1034 		}
1035 		len -= msg.msg_controllen;
1036 		return put_user(len, optlen);
1037 	}
1038 	case IPV6_MTU:
1039 	{
1040 		struct dst_entry *dst;
1041 
1042 		val = 0;
1043 		rcu_read_lock();
1044 		dst = __sk_dst_get(sk);
1045 		if (dst)
1046 			val = dst_mtu(dst);
1047 		rcu_read_unlock();
1048 		if (!val)
1049 			return -ENOTCONN;
1050 		break;
1051 	}
1052 
1053 	case IPV6_V6ONLY:
1054 		val = np->ipv6only;
1055 		break;
1056 
1057 	case IPV6_RECVPKTINFO:
1058 		val = np->rxopt.bits.rxinfo;
1059 		break;
1060 
1061 	case IPV6_2292PKTINFO:
1062 		val = np->rxopt.bits.rxoinfo;
1063 		break;
1064 
1065 	case IPV6_RECVHOPLIMIT:
1066 		val = np->rxopt.bits.rxhlim;
1067 		break;
1068 
1069 	case IPV6_2292HOPLIMIT:
1070 		val = np->rxopt.bits.rxohlim;
1071 		break;
1072 
1073 	case IPV6_RECVRTHDR:
1074 		val = np->rxopt.bits.srcrt;
1075 		break;
1076 
1077 	case IPV6_2292RTHDR:
1078 		val = np->rxopt.bits.osrcrt;
1079 		break;
1080 
1081 	case IPV6_HOPOPTS:
1082 	case IPV6_RTHDRDSTOPTS:
1083 	case IPV6_RTHDR:
1084 	case IPV6_DSTOPTS:
1085 	{
1086 
1087 		lock_sock(sk);
1088 		len = ipv6_getsockopt_sticky(sk, np->opt,
1089 					     optname, optval, len);
1090 		release_sock(sk);
1091 		/* check if ipv6_getsockopt_sticky() returns err code */
1092 		if (len < 0)
1093 			return len;
1094 		return put_user(len, optlen);
1095 	}
1096 
1097 	case IPV6_RECVHOPOPTS:
1098 		val = np->rxopt.bits.hopopts;
1099 		break;
1100 
1101 	case IPV6_2292HOPOPTS:
1102 		val = np->rxopt.bits.ohopopts;
1103 		break;
1104 
1105 	case IPV6_RECVDSTOPTS:
1106 		val = np->rxopt.bits.dstopts;
1107 		break;
1108 
1109 	case IPV6_2292DSTOPTS:
1110 		val = np->rxopt.bits.odstopts;
1111 		break;
1112 
1113 	case IPV6_TCLASS:
1114 		val = np->tclass;
1115 		break;
1116 
1117 	case IPV6_RECVTCLASS:
1118 		val = np->rxopt.bits.rxtclass;
1119 		break;
1120 
1121 	case IPV6_FLOWINFO:
1122 		val = np->rxopt.bits.rxflow;
1123 		break;
1124 
1125 	case IPV6_RECVPATHMTU:
1126 		val = np->rxopt.bits.rxpmtu;
1127 		break;
1128 
1129 	case IPV6_PATHMTU:
1130 	{
1131 		struct dst_entry *dst;
1132 		struct ip6_mtuinfo mtuinfo;
1133 
1134 		if (len < sizeof(mtuinfo))
1135 			return -EINVAL;
1136 
1137 		len = sizeof(mtuinfo);
1138 		memset(&mtuinfo, 0, sizeof(mtuinfo));
1139 
1140 		rcu_read_lock();
1141 		dst = __sk_dst_get(sk);
1142 		if (dst)
1143 			mtuinfo.ip6m_mtu = dst_mtu(dst);
1144 		rcu_read_unlock();
1145 		if (!mtuinfo.ip6m_mtu)
1146 			return -ENOTCONN;
1147 
1148 		if (put_user(len, optlen))
1149 			return -EFAULT;
1150 		if (copy_to_user(optval, &mtuinfo, len))
1151 			return -EFAULT;
1152 
1153 		return 0;
1154 		break;
1155 	}
1156 
1157 	case IPV6_TRANSPARENT:
1158 		val = inet_sk(sk)->transparent;
1159 		break;
1160 
1161 	case IPV6_RECVORIGDSTADDR:
1162 		val = np->rxopt.bits.rxorigdstaddr;
1163 		break;
1164 
1165 	case IPV6_UNICAST_HOPS:
1166 	case IPV6_MULTICAST_HOPS:
1167 	{
1168 		struct dst_entry *dst;
1169 
1170 		if (optname == IPV6_UNICAST_HOPS)
1171 			val = np->hop_limit;
1172 		else
1173 			val = np->mcast_hops;
1174 
1175 		if (val < 0) {
1176 			rcu_read_lock();
1177 			dst = __sk_dst_get(sk);
1178 			if (dst)
1179 				val = ip6_dst_hoplimit(dst);
1180 			rcu_read_unlock();
1181 		}
1182 
1183 		if (val < 0)
1184 			val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1185 		break;
1186 	}
1187 
1188 	case IPV6_MULTICAST_LOOP:
1189 		val = np->mc_loop;
1190 		break;
1191 
1192 	case IPV6_MULTICAST_IF:
1193 		val = np->mcast_oif;
1194 		break;
1195 
1196 	case IPV6_UNICAST_IF:
1197 		val = (__force int)htonl((__u32) np->ucast_oif);
1198 		break;
1199 
1200 	case IPV6_MTU_DISCOVER:
1201 		val = np->pmtudisc;
1202 		break;
1203 
1204 	case IPV6_RECVERR:
1205 		val = np->recverr;
1206 		break;
1207 
1208 	case IPV6_FLOWINFO_SEND:
1209 		val = np->sndflow;
1210 		break;
1211 
1212 	case IPV6_ADDR_PREFERENCES:
1213 		val = 0;
1214 
1215 		if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1216 			val |= IPV6_PREFER_SRC_TMP;
1217 		else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1218 			val |= IPV6_PREFER_SRC_PUBLIC;
1219 		else {
1220 			/* XXX: should we return system default? */
1221 			val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1222 		}
1223 
1224 		if (np->srcprefs & IPV6_PREFER_SRC_COA)
1225 			val |= IPV6_PREFER_SRC_COA;
1226 		else
1227 			val |= IPV6_PREFER_SRC_HOME;
1228 		break;
1229 
1230 	case IPV6_MINHOPCOUNT:
1231 		val = np->min_hopcount;
1232 		break;
1233 
1234 	case IPV6_DONTFRAG:
1235 		val = np->dontfrag;
1236 		break;
1237 
1238 	default:
1239 		return -ENOPROTOOPT;
1240 	}
1241 	len = min_t(unsigned int, sizeof(int), len);
1242 	if(put_user(len, optlen))
1243 		return -EFAULT;
1244 	if(copy_to_user(optval,&val,len))
1245 		return -EFAULT;
1246 	return 0;
1247 }
1248 
1249 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1250 		    char __user *optval, int __user *optlen)
1251 {
1252 	int err;
1253 
1254 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1255 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1256 
1257 	if(level != SOL_IPV6)
1258 		return -ENOPROTOOPT;
1259 
1260 	err = do_ipv6_getsockopt(sk, level, optname, optval, optlen, 0);
1261 #ifdef CONFIG_NETFILTER
1262 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1263 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1264 		int len;
1265 
1266 		if (get_user(len, optlen))
1267 			return -EFAULT;
1268 
1269 		lock_sock(sk);
1270 		err = nf_getsockopt(sk, PF_INET6, optname, optval,
1271 				&len);
1272 		release_sock(sk);
1273 		if (err >= 0)
1274 			err = put_user(len, optlen);
1275 	}
1276 #endif
1277 	return err;
1278 }
1279 
1280 EXPORT_SYMBOL(ipv6_getsockopt);
1281 
1282 #ifdef CONFIG_COMPAT
1283 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1284 			   char __user *optval, int __user *optlen)
1285 {
1286 	int err;
1287 
1288 	if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1289 		if (udp_prot.compat_getsockopt != NULL)
1290 			return udp_prot.compat_getsockopt(sk, level, optname,
1291 							  optval, optlen);
1292 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1293 	}
1294 
1295 	if (level != SOL_IPV6)
1296 		return -ENOPROTOOPT;
1297 
1298 	if (optname == MCAST_MSFILTER)
1299 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1300 			ipv6_getsockopt);
1301 
1302 	err = do_ipv6_getsockopt(sk, level, optname, optval, optlen,
1303 				 MSG_CMSG_COMPAT);
1304 #ifdef CONFIG_NETFILTER
1305 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1306 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1307 		int len;
1308 
1309 		if (get_user(len, optlen))
1310 			return -EFAULT;
1311 
1312 		lock_sock(sk);
1313 		err = compat_nf_getsockopt(sk, PF_INET6,
1314 					   optname, optval, &len);
1315 		release_sock(sk);
1316 		if (err >= 0)
1317 			err = put_user(len, optlen);
1318 	}
1319 #endif
1320 	return err;
1321 }
1322 
1323 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1324 #endif
1325 
1326